Main Session
Sep 29
SS 36 - Predicting Benefit and Toxicity: Biomarkers in Prostate Radiotherapy

313 - Digital Pathology-Based Multimodal Artificial Intelligence Biomarker Validation In Patients with High and Very High Risk Localized Prostate Cancer from the POP-RT Trial

04:05pm - 04:15pm ET
Room 210

Presenter(s)

Vedang Murthy, MD - Tata Memorial Centre, New Mumbai, Maharashtra

V. Murthy1, P. Maitre2, S. Rane3, X. Yang4, M. Tierney4, R. Yamashita4, A. Kraft4, T. N. Showalter5, E. Stewart4, P. Singh2, S. Dubey6, G. Prakash2, R. Krishnatry7, U. M. Mahantshetty8, and S. Menon9; 1ACTREC, Tata Memorial Centre, Homi Bhabha National Institute, Mumbai, India, 2Tata Memorial Centre, Homi Bhabha National Institute, Mumbai, India, 3Pathology and Digital & Computational Oncology, Tata Memorial Centre, Homi Bhabha National Institute, Mumbai, India, 4Artera, Los Altos, CA, 5University of Virginia, Charlottesville, VA, 6Radiation Oncology, Tata Memorial Centre, Homi Bhabha National Institute, Mumbai, India, 7Department of Radiation Oncology, Tata Memorial Centre, Homi Bhabha National Institute, Mumbai, India, 8Homi Bhabha Cancer Hospital and Research Centre, Visakhapatnam, India, 9Tata Memorial Hospital, Tata Memorial Centre, Homi Bhabha National Institute, Mumbai, India

Purpose/Objective(s): Whole-pelvic radiotherapy (WPRT) improves disease control in selected patients with high-risk localized prostate cancer but may overtreat others. We evaluated whether a clinically available digital pathology-based multimodal artificial intelligence (MMAI) biomarker could refine patient selection for WPRT versus prostate-only radiotherapy (PORT) in the randomized phase III POP-RT trial.

Materials/Methods: Pretreatment H&E-stained biopsy slides from POP-RT were digitized and analyzed using the MMAI biomarker, integrating histopathologic features with baseline clinical variables. MMAI scores were assessed continuously and categorically. Associations with distant metastasis (DM), biochemical failure (BF), disease-free survival (DFS), and overall survival (OS) were analyzed using Fine–Gray and Cox proportional hazards models. Outcomes were compared between WPRT and PORT within MMAI risk groups.

Results: Among 173 evaluable patients (91 PORT; 82 WPRT), 55% were NCCN Very High Risk, 50.3% were GG 4-5 and 80% were T3-4. Overall, 110 (64%) were classified as MMAI-high risk and 63 (36%) as MMAI-intermediate risk. MMAI scores were independently prognostic for all endpoints in univariable analysis and for DM (Table 1), BF, and DFS in multivariable analysis, controlling for key clinical variables. After a median follow up of 5.4 years, WPRT significantly reduced 5-year DM and BF among MMAI-high patients compared with PORT (DM: 2% vs 17%, P=0.02; BF: 4% vs 22%, P=0.003). In contrast, no statistically significant differences were observed in MMAI-intermediate patients (DM: 3% vs 0%, P=0.33; BF: 3% vs 14%, P=0.26). Treatment–biomarker interaction was not statistically significant, consistent with limited power.

Conclusion: In a cohort of primarily PSMA PET-staged NCCN high- and very high-risk patients with prostate cancer from the POP-RT trial, a digital pathology-based MMAI biomarker was prognostic for DM, BF and DFS and identified patients most likely to benefit from elective pelvic irradiation.

Abstract 313 - Table 1: MMAI Prognostic Performance for DM, Adjusted for Clinical Variables

  1. Per SD.
Clinical Variable

 MMAI1 sHR (95% CI)

 p-value

 Clinical Variable sHR (95% CI)

 p-value
Age

 1.71 (1.19-2.44)

 0.003*

 0.98 (0.92-1.04)

 0.49
Treatment

(WPRT vs PORT)

 1.66 (1.14-2.41)

 0.008*

 0.29 (0.08-1.01)

 0.05
PSA

(>50 vs =50)

 1.69 (1.16-2.47)

 0.007*

 1.27 (0.44-3.62)

 0.66
NCCN risk group

(Very high vs high)

 1.71 (1.13-2.58)

 0.01*

 1.25 (0.38-4.16)

 0.71
Grade Group

(4-5 vs 1-3)

 1.59 (1.07-2.36)

 0.02*

 1.98 (0.61-6.38)

 0.25
Clinical T-stage

(=T3b vs <T3b)

 1.71 (1.13-2.58)

 0.01*

 1.23 (0.39-3.89)

 0.73
Roach Nodal Risk

(>40% vs =40%)

 1.66 (1.13-2.43)

 0.009*

 1.49 (0.52-4.30)

 0.46